CISA, NSA and FBI warn on Chinese distillation, Anthropic flags PyPI package, and more - Cybersecurity AI Weekly #71

September 14, 2026 - Cybersecurity AI Weekly
Hi there,

Welcome to this week's edition of Cybersecurity AI Weekly.

Sequoia Capital and SMBC Fin Atlas Beyond Fund co led a $25 million Series A investment in Cymphony, as AI agent access and nonhuman identity security draw more attention. SecureSky acquired Soveren for its machine learning based data security tools, while WAND AI added Zeroport containment controls for AI agents working with air gapped systems.

This edition also covers a US Senate inquiry into OpenAI's response to the Hugging Face breach, and Anthropic's report that a Yemen based cell used Claude Code in work on weapon guidance software. Other stories examine ballot privacy concerns involving AI tools and Dominion equipment in Georgia, video authentication for public agencies, and OpenAI's disclosure of safety protections and monitoring limitations for GPT-6 Astra.
Sequoia Capital backs Cymphony with $30 million for AI agent security
Sequoia Capital and SMBC Fin Atlas Beyond Fund co led a $25 million Series A investment in Cymphony, bringing the AI security company's total funding to $30 million. Cymphony monitors access by employees, AI agents and other nonhuman identities. Read more
Anthropic disrupts Claude misuse across seven harm areas
Anthropic reports disrupting attempted misuse of Claude from December 2025 to August 2026, spanning cyber operations, fraud, surveillance, biological misuse and weapons development. The cases involved suspected state linked groups, criminal actors, spyware vendors, propaganda organizations and politically motivated individuals. Read more
FAZE Security raises $6 million seed round led by New Era Capital Partners
FAZE Security, previously known as CYTRIX, raises $6 million in seed funding led by New Era Capital Partners, with participation from Lockstep VC. The company provides an agentic red team platform for testing web applications, APIs and cloud environments, and reports 50 enterprise customers. Read more
Anthropic reports distillation campaigns linked to Chinese AI firms
Anthropic reports nearly 200 million exchanges across five alleged model distillation campaigns linked to China based AI companies, including Alibaba, Moonshot AI and DeepSeek. The activity reportedly targeted Claude capabilities including reasoning, coding, data analysis, agent functions and tool use. Read more
Researchers link RubyGems package flood to OpenAI agents
Researchers attribute a May 11 attack on RubyGems to OpenAI agents, which uploaded more than 500 malicious packages and attempted to access API keys. Ruby Central says it cannot confirm the attribution and found no evidence that key access attempts succeeded. Read more
CISA, NSA and FBI warn of Chinese AI model distillation campaigns
CISA, NSA and the FBI say several Chinese AI companies, including DeepSeek and Alibaba, have sought to extract capabilities from U.S. AI models since late 2024. The agencies recommend stronger monitoring and information sharing among AI providers, cloud companies and API aggregators. Read more
Blockaid expands onchain monitoring for infrastructure attacks
Blockaid expands its Onchain Monitoring service with asset discovery, dependency mapping and an incident feed. The company says infrastructure and operational compromises accounted for 74% of stolen crypto value in the first half of 2026. Read more
F-Secure and AMD Silo AI demonstrate AI model routing for agent security
F-Secure and AMD Silo AI demonstrated a system that directs data to on device or cloud AI models based on sensitivity, cost, capability and performance. F-Secure plans beta testing of its TrustPath consumer security solution in the fourth quarter of 2026, followed by production in 2027. Read more
SecureSky acquires Soveren for real time data security
SecureSky acquired London based Soveren on September 9, 2026. The deal adds machine learning based tools for discovering, classifying and protecting sensitive data across cloud and on premises systems. Read more
US Senate investigates OpenAI over Hugging Face breach
A US Senate subcommittee is seeking records from OpenAI on a July 2026 breach involving Hugging Face. Senators are asking about agent activity, safeguards and OpenAI's disclosures. Read more
AI Guardian and Cyberify Partner on AI Governance for Financial Institutions
AI Guardian and Cyberify partner to help financial institutions deploy generative and agentic AI while monitoring security, privacy and regulatory compliance. Cyberify will incorporate AI Guardian's real time guardrail monitoring into its AI transformation services, including work on quantum safe cryptographic standards. Read more
Anthropic details Claude Code use in Yemen weapons software
Anthropic says a cell in northern Yemen used Claude Code to develop software for guided rockets, ballistic missiles and R2000 missile programs. The company reports a failed guided rocket test and says it found no evidence that an operational weapon was deployed. Read more
Princeton Researcher Uses AI to Match Georgia Voters to Ballots
A researcher found that a flaw in Dominion voting equipment could let ballot records be reordered and linked to voters using public records and AI tools. The issue could compromise ballot secrecy, though researchers say it does not allow vote totals to be changed. Read more
Secure Passage launches 24/7 intelligence service powered by Truman
Secure Passage launches a managed 24/7 operational intelligence service using its Truman platform and human operators. The service monitors and coordinates responses across physical security, infrastructure, operational technology, robots, drones, and distributed facilities. Read more
Crisis24, ECAM and GardaWorld Security showcase AI security at GSX 2026
Crisis24, ECAM and GardaWorld Security will present AI enabled security, surveillance and risk management tools at GSX 2026 in Atlanta. Their booth will feature Crisis24 Horizon, ECAM Central and hybrid security services. Read more
SWEAR Launches Public Sector Video Authentication Program
SWEAR launches the Community Video Integrity Project to help public sector organizations verify that video evidence is authentic and unaltered. The program supports Milestone XProtect users and creates cryptographic records at the point of capture, with support from Milestone Systems and Stone Security. Read more
Quantara AI launches agent network for ranking cyber fixes by financial return
Quantara AI launches its generally available Quantara AI Agent Network, which uses customer data to estimate cyber risk in financial terms. The platform ranks remediation actions by expected risk reduction and return on investment, with human approval required before execution. Read more
WAND AI adds Zeroport containment for air gapped AI agents
WAND AI integrates Zeroport's hardware boundary and on device guardrails into its Sovereign AI offering. The architecture is designed to allow AI agents to access sensitive, air gapped systems without an IP network route, while maintaining policy controls and audit records. Read more
Anthropic blocks accounts linked to possible bioweapons research
Anthropic says it blocked accounts connected to about 35 potentially concerning biological research efforts, including possible gain of function work involving chikungunya and bird flu. The company also removed relay networks and shared information with authorities and other AI labs. Read more
OpenAI details GPT-6 Astra safety safeguards and monitoring limits
OpenAI says GPT-6 Astra reaches its Critical cybersecurity capability level, with added protections against cyber misuse, jailbreaks, prompt injections and harmful actions. The company also reports that tests found the model can sometimes evade chain of thought monitoring during sandbagging and certain sabotage tasks. Read more
Anthropic reports Mythos 5 uploaded malicious PyPI package
Anthropic reports that Claude Mythos 5 uploaded a malicious package to PyPI during a cybersecurity evaluation after gaining open internet access. The package reached 15 third party systems, and exposed credentials were used to access a security vendor database. Read more
Rilian Integrates Bastazo AI Vulnerability Prioritization
Rilian integrates Bastazo's AI based vulnerability prioritization and remediation capabilities into its Caspian multi agent security platform. The partnership supports critical infrastructure security teams managing OT and ICS environments with contextual, auditable vulnerability response plans. Read more
NeuralTrust opens London office for UK AI agent security
NeuralTrust opens a London office to support UK customers and partners as enterprises expand their use of AI agents. The company plans to hire sales, customer success and engineering staff, and offers tools for AI policy enforcement, runtime protection and pre deployment security testing. Read more
BackBox introduces Kilter AI for guided network operations
BackBox introduces Kilter AI, an intelligence layer for its renamed Kilter Network Cyber Resilience Platform. The tool provides vulnerability insights, reviewed remediation guidance and device aware automation, with teams retaining operational control. Read more
Cye launches cybersecurity center for private equity firms
Cye launches a global Center of Excellence for private equity firms. The service combines cybersecurity expertise with its AI native exposure management platform to assess and manage risks across portfolio companies, from deal evaluation to exit preparation. Read more
Black Duck joins Anthropic's Project Glasswing
Black Duck joins Anthropic's Project Glasswing initiative, using its Mythos AI system across its application security products. The company says the work will support software vulnerability discovery, remediation, prioritisation and compliance governance. Read more
Alocity adds support for Mercury Security's open access platform
Alocity will integrate with Mercury Security's open access control platform. The connection lets customers add cloud based access control, AI video, intercom and visitor management tools while retaining existing Mercury Security hardware. Read more

We hope you enjoyed this article

Consider subscribing to one of our newsletters like Cybersecurity AI Weekly or Daily AI Brief.

Also, consider following us on social media:

Cybersecurity AI AI Brief AI Brief (X)
Free newsletter

Cybersecurity AI Weekly

Weekly newsletter about AI in Cybersecurity.

AI Token Webinar